Aspire predicts and refines atomic rearrangements faster and more completely
ASPIRE: Saddle-Point Discovery through Set Prediction and Physical Refinement
Machine Learning
Summary
Finding how atoms move and rearrange themselves is important for understanding materials but can be very slow and tricky. The authors created ASPIRE, a system that predicts several possible atomic moves from one setup and then refines each prediction to find exact transition points more efficiently. They also built new large datasets to help train and test such predictions. Their method covers more real atomic moves than a standard method while using less computing time, meaning it can help study atomic behaviors faster.
What this means in practice
- •For materials simulation teams: Identify and refine multiple atomic rearrangement mechanisms in materials more efficiently to accelerate simulation of defect evolution.
- •For computational chemists: Generate diverse and accurate saddle-point candidates for chemical reaction pathways to reduce computational costs.
Authors
Yucheng Zhao, Quanyou Zhang, Shaoxiang Qin, Haixuan Xu, Xiongye Xiao
Abstract
Predicting thermally activated diffusion and defect evolution with event-driven models requires identifying atomic rearrangement mechanisms and their activation barriers. Discovering the associated saddle points is a major computational bottleneck: multiple rearrangements may originate from one metastable state, while costly local searches can fail or repeatedly converge to the same saddle. To address this challenge, we introduce ASPIRE (Atomistic Saddle-Point Inference with Refinement for Events), a framework that predicts a set of saddle candidates from a single initial atomic environment and refines them through Dimer searches on the original interatomic potential. The framework's equivariant set predictor, Ev-Quiformer, integrates (i) geometry-conditioned scalar-vector event slots for generating multiple saddle-point proposals and (ii) a decoder that maps each slot to a full atomic displacement field by combining atom, slot, and anchor-relative vectors with invariant coefficients. We also contribute two datasets: (i) BCCFE4VACAV-4000, comprising 4,000 four-vacancy body-centered cubic iron configurations and 65,450 reference events grouped by initial state for set supervision and post-refinement evaluation; and (ii) BCCFE-1TO4VAC, comprising 5,372 configurations with one to four vacancies each. Theoretically, we establish conditions for proposal equivariance. Experimentally, ASPIRE achieves 77.20% reference-event coverage on this benchmark, compared with 75.73% for a conventional Dimer baseline, while requiring approximately half as many Dimer force evaluations. In a timing evaluation on 50 configurations, ASPIRE reduces wall time per configuration from 478.8 s to 176.3 s under the stated hardware settings.